1  ---
  2  title: "IBM watsonx.ai"
  3  id: integrations-watsonx
  4  description: "IBM watsonx.ai integration for Haystack"
  5  slug: "/integrations-watsonx"
  6  ---
  7  
  8  
  9  ## haystack_integrations.components.embedders.watsonx.document_embedder
 10  
 11  ### WatsonxDocumentEmbedder
 12  
 13  Computes document embeddings using IBM watsonx.ai models.
 14  
 15  ### Usage example
 16  
 17  ```python
 18  from haystack import Document
 19  from haystack_integrations.components.embedders.watsonx.document_embedder import WatsonxDocumentEmbedder
 20  
 21  documents = [
 22      Document(content="I love pizza!"),
 23      Document(content="Pasta is great too"),
 24  ]
 25  
 26  document_embedder = WatsonxDocumentEmbedder(
 27      model="ibm/slate-30m-english-rtrvr-v2",
 28      api_key=Secret.from_env_var("WATSONX_API_KEY"),
 29      api_base_url="https://us-south.ml.cloud.ibm.com",
 30      project_id=Secret.from_env_var("WATSONX_PROJECT_ID"),
 31  )
 32  
 33  result = document_embedder.run(documents=documents)
 34  print(result["documents"][0].embedding)
 35  
 36  # [0.017020374536514282, -0.023255806416273117, ...]
 37  ```
 38  
 39  #### __init__
 40  
 41  ```python
 42  __init__(
 43      *,
 44      model: str = "ibm/slate-30m-english-rtrvr-v2",
 45      api_key: Secret = Secret.from_env_var("WATSONX_API_KEY"),
 46      api_base_url: str = "https://us-south.ml.cloud.ibm.com",
 47      project_id: Secret = Secret.from_env_var("WATSONX_PROJECT_ID"),
 48      truncate_input_tokens: int | None = None,
 49      prefix: str = "",
 50      suffix: str = "",
 51      batch_size: int = 1000,
 52      concurrency_limit: int = 5,
 53      timeout: float | None = None,
 54      max_retries: int | None = None,
 55      meta_fields_to_embed: list[str] | None = None,
 56      embedding_separator: str = "\n"
 57  )
 58  ```
 59  
 60  Creates a WatsonxDocumentEmbedder component.
 61  
 62  **Parameters:**
 63  
 64  - **model** (<code>str</code>) – The name of the model to use for calculating embeddings.
 65    Default is "ibm/slate-30m-english-rtrvr-v2".
 66  - **api_key** (<code>Secret</code>) – The WATSONX API key. Can be set via environment variable WATSONX_API_KEY.
 67  - **api_base_url** (<code>str</code>) – The WATSONX URL for the watsonx.ai service.
 68    Default is "https://us-south.ml.cloud.ibm.com".
 69  - **project_id** (<code>Secret</code>) – The ID of the Watson Studio project.
 70    Can be set via environment variable WATSONX_PROJECT_ID.
 71  - **truncate_input_tokens** (<code>int | None</code>) – Maximum number of tokens to use from the input text.
 72    If set to `None` (or not provided), the full input text is used, up to the model's maximum token limit.
 73  - **prefix** (<code>str</code>) – A string to add at the beginning of each text.
 74  - **suffix** (<code>str</code>) – A string to add at the end of each text.
 75  - **batch_size** (<code>int</code>) – Number of documents to embed in one API call. Default is 1000.
 76  - **concurrency_limit** (<code>int</code>) – Number of parallel requests to make. Default is 5.
 77  - **timeout** (<code>float | None</code>) – Timeout for API requests in seconds.
 78  - **max_retries** (<code>int | None</code>) – Maximum number of retries for API requests.
 79  
 80  #### to_dict
 81  
 82  ```python
 83  to_dict() -> dict[str, Any]
 84  ```
 85  
 86  Serialize the component to a dictionary.
 87  
 88  **Returns:**
 89  
 90  - <code>dict\[str, Any\]</code> – The serialized component as a dictionary.
 91  
 92  #### from_dict
 93  
 94  ```python
 95  from_dict(data: dict[str, Any]) -> 'WatsonxDocumentEmbedder'
 96  ```
 97  
 98  Deserializes the component from a dictionary.
 99  
100  **Parameters:**
101  
102  - **data** (<code>dict\[str, Any\]</code>) – The dictionary representation of this component.
103  
104  **Returns:**
105  
106  - <code>'WatsonxDocumentEmbedder'</code> – The deserialized component instance.
107  
108  #### run
109  
110  ```python
111  run(documents: list[Document]) -> dict[str, list[Document] | dict[str, Any]]
112  ```
113  
114  Embeds a list of documents.
115  
116  **Parameters:**
117  
118  - **documents** (<code>list\[Document\]</code>) – A list of documents to embed.
119  
120  **Returns:**
121  
122  - <code>dict\[str, list\[Document\] | dict\[str, Any\]\]</code> – A dictionary with:
123  - 'documents': List of Documents with embeddings added
124  - 'meta': Information about the model usage
125  
126  ## haystack_integrations.components.embedders.watsonx.text_embedder
127  
128  ### WatsonxTextEmbedder
129  
130  Embeds strings using IBM watsonx.ai foundation models.
131  
132  You can use it to embed user query and send it to an embedding Retriever.
133  
134  ### Usage example
135  
136  ```python
137  from haystack_integrations.components.embedders.watsonx.text_embedder import WatsonxTextEmbedder
138  
139  text_to_embed = "I love pizza!"
140  
141  text_embedder = WatsonxTextEmbedder(
142      model="ibm/slate-30m-english-rtrvr-v2",
143      api_key=Secret.from_env_var("WATSONX_API_KEY"),
144      api_base_url="https://us-south.ml.cloud.ibm.com",
145      project_id=Secret.from_env_var("WATSONX_PROJECT_ID"),
146  )
147  
148  print(text_embedder.run(text_to_embed))
149  
150  # {'embedding': [0.017020374536514282, -0.023255806416273117, ...],
151  #  'meta': {'model': 'ibm/slate-30m-english-rtrvr-v2',
152  #           'truncated_input_tokens': 3}}
153  ```
154  
155  #### __init__
156  
157  ```python
158  __init__(
159      *,
160      model: str = "ibm/slate-30m-english-rtrvr-v2",
161      api_key: Secret = Secret.from_env_var("WATSONX_API_KEY"),
162      api_base_url: str = "https://us-south.ml.cloud.ibm.com",
163      project_id: Secret = Secret.from_env_var("WATSONX_PROJECT_ID"),
164      truncate_input_tokens: int | None = None,
165      prefix: str = "",
166      suffix: str = "",
167      timeout: float | None = None,
168      max_retries: int | None = None
169  )
170  ```
171  
172  Creates an WatsonxTextEmbedder component.
173  
174  **Parameters:**
175  
176  - **model** (<code>str</code>) – The name of the IBM watsonx model to use for calculating embeddings.
177    Default is "ibm/slate-30m-english-rtrvr-v2".
178  - **api_key** (<code>Secret</code>) – The WATSONX API key. Can be set via environment variable WATSONX_API_KEY.
179  - **api_base_url** (<code>str</code>) – The WATSONX URL for the watsonx.ai service.
180    Default is "https://us-south.ml.cloud.ibm.com".
181  - **project_id** (<code>Secret</code>) – The ID of the Watson Studio project.
182    Can be set via environment variable WATSONX_PROJECT_ID.
183  - **truncate_input_tokens** (<code>int | None</code>) – Maximum number of tokens to use from the input text.
184    If set to `None` (or not provided), the full input text is used, up to the model's maximum token limit.
185  - **prefix** (<code>str</code>) – A string to add at the beginning of each text to embed.
186  - **suffix** (<code>str</code>) – A string to add at the end of each text to embed.
187  - **timeout** (<code>float | None</code>) – Timeout for API requests in seconds.
188  - **max_retries** (<code>int | None</code>) – Maximum number of retries for API requests.
189  
190  #### to_dict
191  
192  ```python
193  to_dict() -> dict[str, Any]
194  ```
195  
196  Serialize the component to a dictionary.
197  
198  **Returns:**
199  
200  - <code>dict\[str, Any\]</code> – The serialized component as a dictionary.
201  
202  #### from_dict
203  
204  ```python
205  from_dict(data: dict[str, Any]) -> WatsonxTextEmbedder
206  ```
207  
208  Deserializes the component from a dictionary.
209  
210  **Parameters:**
211  
212  - **data** (<code>dict\[str, Any\]</code>) – The dictionary representation of this component.
213  
214  **Returns:**
215  
216  - <code>WatsonxTextEmbedder</code> – The deserialized component instance.
217  
218  #### run
219  
220  ```python
221  run(text: str) -> dict[str, list[float] | dict[str, Any]]
222  ```
223  
224  Embeds a single string.
225  
226  **Parameters:**
227  
228  - **text** (<code>str</code>) – Text to embed.
229  
230  **Returns:**
231  
232  - <code>dict\[str, list\[float\] | dict\[str, Any\]\]</code> – A dictionary with:
233  - 'embedding': The embedding of the input text
234  - 'meta': Information about the model usage
235  
236  ## haystack_integrations.components.generators.watsonx.chat.chat_generator
237  
238  ### WatsonxChatGenerator
239  
240  Enables chat completions using IBM's watsonx.ai foundation models.
241  
242  This component interacts with IBM's watsonx.ai platform to generate chat responses using various foundation
243  models. It supports the [ChatMessage](https://docs.haystack.deepset.ai/docs/chatmessage) format for both input
244  and output, including multimodal inputs with text and images.
245  
246  The generator works with IBM's foundation models that are listed
247  [here](https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx&audience=wdp).
248  
249  You can customize the generation behavior by passing parameters to the watsonx.ai API through the
250  `generation_kwargs` argument. These parameters are passed directly to the watsonx.ai inference endpoint.
251  
252  For details on watsonx.ai API parameters, see
253  [IBM watsonx.ai documentation](https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-parameters.html).
254  
255  ### Usage example
256  
257  ```python
258  from haystack_integrations.components.generators.watsonx.chat.chat_generator import WatsonxChatGenerator
259  from haystack.dataclasses import ChatMessage
260  from haystack.utils import Secret
261  
262  messages = [ChatMessage.from_user("Explain quantum computing in simple terms")]
263  
264  client = WatsonxChatGenerator(
265      api_key=Secret.from_env_var("WATSONX_API_KEY"),
266      model="ibm/granite-4-h-small",
267      project_id=Secret.from_env_var("WATSONX_PROJECT_ID"),
268  )
269  response = client.run(messages)
270  print(response)
271  ```
272  
273  ### Multimodal usage example
274  
275  ```python
276  from haystack.dataclasses import ChatMessage, ImageContent
277  
278  # Create an image from file path or base64
279  image_content = ImageContent.from_file_path("path/to/your/image.jpg")
280  
281  # Create a multimodal message with both text and image
282  messages = [ChatMessage.from_user(content_parts=["What's in this image?", image_content])]
283  
284  # Use a multimodal model
285  client = WatsonxChatGenerator(
286      api_key=Secret.from_env_var("WATSONX_API_KEY"),
287      model="meta-llama/llama-3-2-11b-vision-instruct",
288      project_id=Secret.from_env_var("WATSONX_PROJECT_ID"),
289  )
290  response = client.run(messages)
291  print(response)
292  ```
293  
294  #### SUPPORTED_MODELS
295  
296  ```python
297  SUPPORTED_MODELS: list[str] = [
298      "ibm/granite-3-1-8b-base",
299      "ibm/granite-3-8b-instruct",
300      "ibm/granite-4-h-small",
301      "ibm/granite-8b-code-instruct",
302      "ibm/granite-guardian-3-8b",
303      "meta-llama/llama-3-1-70b-gptq",
304      "meta-llama/llama-3-1-8b",
305      "meta-llama/llama-3-2-11b-vision-instruct",
306      "meta-llama/llama-3-2-90b-vision-instruct",
307      "meta-llama/llama-3-3-70b-instruct",
308      "meta-llama/llama-3-405b-instruct",
309      "meta-llama/llama-4-maverick-17b-128e-instruct-fp8",
310      "meta-llama/llama-guard-3-11b-vision",
311      "mistral-large-2512",
312      "mistralai/mistral-medium-2505",
313      "mistralai/mistral-small-3-1-24b-instruct-2503",
314      "openai/gpt-oss-120b",
315  ]
316  
317  ```
318  
319  A non-exhaustive list of models supported by this component.
320  
321  See https://www.ibm.com/docs/en/watsonx/saas?topic=solutions-supported-foundation-models for the
322  full list of models and up-to-date model IDs.
323  
324  #### __init__
325  
326  ```python
327  __init__(
328      *,
329      api_key: Secret = Secret.from_env_var("WATSONX_API_KEY"),
330      model: str = "ibm/granite-4-h-small",
331      project_id: Secret = Secret.from_env_var("WATSONX_PROJECT_ID"),
332      api_base_url: str = "https://us-south.ml.cloud.ibm.com",
333      generation_kwargs: dict[str, Any] | None = None,
334      timeout: float | None = None,
335      max_retries: int | None = None,
336      verify: bool | str | None = None,
337      streaming_callback: StreamingCallbackT | None = None,
338      tools: ToolsType | None = None
339  ) -> None
340  ```
341  
342  Creates an instance of WatsonxChatGenerator.
343  
344  Before initializing the component, you can set environment variables:
345  
346  - `WATSONX_TIMEOUT` to override the default timeout
347  - `WATSONX_MAX_RETRIES` to override the default retry count
348  
349  **Parameters:**
350  
351  - **api_key** (<code>Secret</code>) – IBM Cloud API key for watsonx.ai access.
352    Can be set via `WATSONX_API_KEY` environment variable or passed directly.
353  - **model** (<code>str</code>) – The model ID to use for completions. Defaults to "ibm/granite-4-h-small".
354    Available models can be found in your IBM Cloud account.
355  - **project_id** (<code>Secret</code>) – IBM Cloud project ID
356  - **api_base_url** (<code>str</code>) – Custom base URL for the API endpoint.
357    Defaults to "https://us-south.ml.cloud.ibm.com".
358  - **generation_kwargs** (<code>dict\[str, Any\] | None</code>) – Additional parameters to control text generation.
359    These parameters are passed directly to the watsonx.ai inference endpoint.
360    Supported parameters include:
361  - `temperature`: Controls randomness (lower = more deterministic)
362  - `max_new_tokens`: Maximum number of tokens to generate
363  - `min_new_tokens`: Minimum number of tokens to generate
364  - `top_p`: Nucleus sampling probability threshold
365  - `top_k`: Number of highest probability tokens to consider
366  - `repetition_penalty`: Penalty for repeated tokens
367  - `length_penalty`: Penalty based on output length
368  - `stop_sequences`: List of sequences where generation should stop
369  - `random_seed`: Seed for reproducible results
370  - **timeout** (<code>float | None</code>) – Timeout in seconds for API requests.
371    Defaults to environment variable `WATSONX_TIMEOUT` or 30 seconds.
372  - **max_retries** (<code>int | None</code>) – Maximum number of retry attempts for failed requests.
373    Defaults to environment variable `WATSONX_MAX_RETRIES` or 5.
374  - **verify** (<code>bool | str | None</code>) – SSL verification setting. Can be:
375  - True: Verify SSL certificates (default)
376  - False: Skip verification (insecure)
377  - Path to CA bundle for custom certificates
378  - **streaming_callback** (<code>StreamingCallbackT | None</code>) – A callback function for streaming responses.
379  - **tools** (<code>ToolsType | None</code>) – A list of Tool and/or Toolset objects, or a single Toolset for which the model can prepare calls.
380  
381  #### to_dict
382  
383  ```python
384  to_dict() -> dict[str, Any]
385  ```
386  
387  Serialize the component to a dictionary.
388  
389  **Returns:**
390  
391  - <code>dict\[str, Any\]</code> – The serialized component as a dictionary.
392  
393  #### from_dict
394  
395  ```python
396  from_dict(data: dict[str, Any]) -> WatsonxChatGenerator
397  ```
398  
399  Deserialize this component from a dictionary.
400  
401  **Parameters:**
402  
403  - **data** (<code>dict\[str, Any\]</code>) – The dictionary representation of this component.
404  
405  **Returns:**
406  
407  - <code>WatsonxChatGenerator</code> – The deserialized component instance.
408  
409  #### run
410  
411  ```python
412  run(
413      *,
414      messages: list[ChatMessage],
415      generation_kwargs: dict[str, Any] | None = None,
416      streaming_callback: StreamingCallbackT | None = None,
417      tools: ToolsType | None = None
418  ) -> dict[str, list[ChatMessage]]
419  ```
420  
421  Generate chat completions synchronously.
422  
423  **Parameters:**
424  
425  - **messages** (<code>list\[ChatMessage\]</code>) – A list of ChatMessage instances representing the input messages.
426  - **generation_kwargs** (<code>dict\[str, Any\] | None</code>) – Additional keyword arguments for text generation. These parameters will potentially override the parameters
427    passed in the `__init__` method.
428  - **streaming_callback** (<code>StreamingCallbackT | None</code>) – A callback function that is called when a new token is received from the stream.
429    If provided this will override the `streaming_callback` set in the `__init__` method.
430  - **tools** (<code>ToolsType | None</code>) – A list of Tool and/or Toolset objects, or a single Toolset for which the model can prepare calls.
431    If set, it will override the `tools` parameter provided during initialization.
432  
433  **Returns:**
434  
435  - <code>dict\[str, list\[ChatMessage\]\]</code> – A dictionary with the following key:
436  - `replies`: A list containing the generated responses as ChatMessage instances.
437  
438  #### run_async
439  
440  ```python
441  run_async(
442      *,
443      messages: list[ChatMessage],
444      generation_kwargs: dict[str, Any] | None = None,
445      streaming_callback: StreamingCallbackT | None = None,
446      tools: ToolsType | None = None
447  ) -> dict[str, list[ChatMessage]]
448  ```
449  
450  Generate chat completions asynchronously.
451  
452  **Parameters:**
453  
454  - **messages** (<code>list\[ChatMessage\]</code>) – A list of ChatMessage instances representing the input messages.
455  - **generation_kwargs** (<code>dict\[str, Any\] | None</code>) – Additional keyword arguments for text generation. These parameters will potentially override the parameters
456    passed in the `__init__` method.
457  - **streaming_callback** (<code>StreamingCallbackT | None</code>) – A callback function that is called when a new token is received from the stream.
458    If provided this will override the `streaming_callback` set in the `__init__` method.
459  - **tools** (<code>ToolsType | None</code>) – A list of Tool and/or Toolset objects, or a single Toolset for which the model can prepare calls.
460    If set, it will override the `tools` parameter provided during initialization.
461  
462  **Returns:**
463  
464  - <code>dict\[str, list\[ChatMessage\]\]</code> – A dictionary with the following key:
465  - `replies`: A list containing the generated responses as ChatMessage instances.
466  
467  ## haystack_integrations.components.generators.watsonx.generator
468  
469  ### WatsonxGenerator
470  
471  Bases: <code>WatsonxChatGenerator</code>
472  
473  Enables text completions using IBM's watsonx.ai foundation models.
474  
475  This component extends WatsonxChatGenerator to provide the standard Generator interface that works with prompt
476  strings instead of ChatMessage objects.
477  
478  The generator works with IBM's foundation models that are listed
479  [here](https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx&audience=wdp).
480  
481  You can customize the generation behavior by passing parameters to the watsonx.ai API through the
482  `generation_kwargs` argument. These parameters are passed directly to the watsonx.ai inference endpoint.
483  
484  For details on watsonx.ai API parameters, see
485  [IBM watsonx.ai documentation](https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-parameters.html).
486  
487  ### Usage example
488  
489  ```python
490  from haystack_integrations.components.generators.watsonx.generator import WatsonxGenerator
491  from haystack.utils import Secret
492  
493  generator = WatsonxGenerator(
494      api_key=Secret.from_env_var("WATSONX_API_KEY"),
495      model="ibm/granite-4-h-small",
496      project_id=Secret.from_env_var("WATSONX_PROJECT_ID"),
497  )
498  
499  response = generator.run(
500      prompt="Explain quantum computing in simple terms",
501      system_prompt="You are a helpful physics teacher.",
502  )
503  print(response)
504  ```
505  
506  Output:
507  
508  ```
509  {
510      "replies": ["Quantum computing uses quantum-mechanical phenomena like...."],
511      "meta": [
512          {
513              "model": "ibm/granite-4-h-small",
514              "project_id": "your-project-id",
515              "usage": {
516                  "prompt_tokens": 12,
517                  "completion_tokens": 45,
518                  "total_tokens": 57,
519              },
520          }
521      ],
522  }
523  ```
524  
525  #### SUPPORTED_MODELS
526  
527  ```python
528  SUPPORTED_MODELS: list[str] = [
529      "ibm/granite-3-1-8b-base",
530      "ibm/granite-3-8b-instruct",
531      "ibm/granite-4-h-small",
532      "ibm/granite-8b-code-instruct",
533      "ibm/granite-guardian-3-8b",
534      "meta-llama/llama-3-1-70b-gptq",
535      "meta-llama/llama-3-1-8b",
536      "meta-llama/llama-3-2-11b-vision-instruct",
537      "meta-llama/llama-3-2-90b-vision-instruct",
538      "meta-llama/llama-3-3-70b-instruct",
539      "meta-llama/llama-3-405b-instruct",
540      "meta-llama/llama-4-maverick-17b-128e-instruct-fp8",
541      "meta-llama/llama-guard-3-11b-vision",
542      "mistral-large-2512",
543      "mistralai/mistral-medium-2505",
544      "mistralai/mistral-small-3-1-24b-instruct-2503",
545      "openai/gpt-oss-120b",
546  ]
547  
548  ```
549  
550  A non-exhaustive list of models supported by this component.
551  
552  See https://www.ibm.com/docs/en/watsonx/saas?topic=solutions-supported-foundation-models for the
553  full list of models and up-to-date model IDs.
554  
555  #### __init__
556  
557  ```python
558  __init__(
559      *,
560      api_key: Secret = Secret.from_env_var("WATSONX_API_KEY"),
561      model: str = "ibm/granite-4-h-small",
562      project_id: Secret = Secret.from_env_var("WATSONX_PROJECT_ID"),
563      api_base_url: str = "https://us-south.ml.cloud.ibm.com",
564      system_prompt: str | None = None,
565      generation_kwargs: dict[str, Any] | None = None,
566      timeout: float | None = None,
567      max_retries: int | None = None,
568      verify: bool | str | None = None,
569      streaming_callback: StreamingCallbackT | None = None
570  ) -> None
571  ```
572  
573  Creates an instance of WatsonxGenerator.
574  
575  Before initializing the component, you can set environment variables:
576  
577  - `WATSONX_TIMEOUT` to override the default timeout
578  - `WATSONX_MAX_RETRIES` to override the default retry count
579  
580  **Parameters:**
581  
582  - **api_key** (<code>Secret</code>) – IBM Cloud API key for watsonx.ai access.
583    Can be set via `WATSONX_API_KEY` environment variable or passed directly.
584  - **model** (<code>str</code>) – The model ID to use for completions. Defaults to "ibm/granite-4-h-small".
585    Available models can be found in your IBM Cloud account.
586  - **project_id** (<code>Secret</code>) – IBM Cloud project ID
587  - **api_base_url** (<code>str</code>) – Custom base URL for the API endpoint.
588    Defaults to "https://us-south.ml.cloud.ibm.com".
589  - **system_prompt** (<code>str | None</code>) – The system prompt to use for text generation.
590  - **generation_kwargs** (<code>dict\[str, Any\] | None</code>) – Additional parameters to control text generation.
591    These parameters are passed directly to the watsonx.ai inference endpoint.
592    Supported parameters include:
593  - `temperature`: Controls randomness (lower = more deterministic)
594  - `max_new_tokens`: Maximum number of tokens to generate
595  - `min_new_tokens`: Minimum number of tokens to generate
596  - `top_p`: Nucleus sampling probability threshold
597  - `top_k`: Number of highest probability tokens to consider
598  - `repetition_penalty`: Penalty for repeated tokens
599  - `length_penalty`: Penalty based on output length
600  - `stop_sequences`: List of sequences where generation should stop
601  - `random_seed`: Seed for reproducible results
602  - **timeout** (<code>float | None</code>) – Timeout in seconds for API requests.
603    Defaults to environment variable `WATSONX_TIMEOUT` or 30 seconds.
604  - **max_retries** (<code>int | None</code>) – Maximum number of retry attempts for failed requests.
605    Defaults to environment variable `WATSONX_MAX_RETRIES` or 5.
606  - **verify** (<code>bool | str | None</code>) – SSL verification setting. Can be:
607  - True: Verify SSL certificates (default)
608  - False: Skip verification (insecure)
609  - Path to CA bundle for custom certificates
610  - **streaming_callback** (<code>StreamingCallbackT | None</code>) – A callback function for streaming responses.
611  
612  #### to_dict
613  
614  ```python
615  to_dict() -> dict[str, Any]
616  ```
617  
618  Serialize the component to a dictionary.
619  
620  **Returns:**
621  
622  - <code>dict\[str, Any\]</code> – The serialized component as a dictionary.
623  
624  #### from_dict
625  
626  ```python
627  from_dict(data: dict[str, Any]) -> WatsonxGenerator
628  ```
629  
630  Deserialize this component from a dictionary.
631  
632  **Parameters:**
633  
634  - **data** (<code>dict\[str, Any\]</code>) – The dictionary representation of this component.
635  
636  **Returns:**
637  
638  - <code>WatsonxGenerator</code> – The deserialized component instance.
639  
640  #### run
641  
642  ```python
643  run(
644      *,
645      prompt: str,
646      system_prompt: str | None = None,
647      streaming_callback: StreamingCallbackT | None = None,
648      generation_kwargs: dict[str, Any] | None = None
649  ) -> dict[str, Any]
650  ```
651  
652  Generate text completions synchronously.
653  
654  **Parameters:**
655  
656  - **prompt** (<code>str</code>) – The input prompt string for text generation.
657  - **system_prompt** (<code>str | None</code>) – An optional system prompt to provide context or instructions for the generation.
658    If not provided, the system prompt set in the `__init__` method will be used.
659  - **streaming_callback** (<code>StreamingCallbackT | None</code>) – A callback function that is called when a new token is received from the stream.
660    If provided, this will override the `streaming_callback` set in the `__init__` method.
661  - **generation_kwargs** (<code>dict\[str, Any\] | None</code>) – Additional keyword arguments for text generation. These parameters will potentially override the parameters
662    passed in the `__init__` method. Supported parameters include temperature, max_new_tokens, top_p, etc.
663  
664  **Returns:**
665  
666  - <code>dict\[str, Any\]</code> – A dictionary with the following keys:
667  - `replies`: A list of generated text completions as strings.
668  - `meta`: A list of metadata dictionaries containing information about each generation,
669    including model name, finish reason, and token usage statistics.
670  
671  #### run_async
672  
673  ```python
674  run_async(
675      *,
676      prompt: str,
677      system_prompt: str | None = None,
678      streaming_callback: StreamingCallbackT | None = None,
679      generation_kwargs: dict[str, Any] | None = None
680  ) -> dict[str, Any]
681  ```
682  
683  Generate text completions asynchronously.
684  
685  **Parameters:**
686  
687  - **prompt** (<code>str</code>) – The input prompt string for text generation.
688  - **system_prompt** (<code>str | None</code>) – An optional system prompt to provide context or instructions for the generation.
689  - **streaming_callback** (<code>StreamingCallbackT | None</code>) – A callback function that is called when a new token is received from the stream.
690    If provided, this will override the `streaming_callback` set in the `__init__` method.
691  - **generation_kwargs** (<code>dict\[str, Any\] | None</code>) – Additional keyword arguments for text generation. These parameters will potentially override the parameters
692    passed in the `__init__` method. Supported parameters include temperature, max_new_tokens, top_p, etc.
693  
694  **Returns:**
695  
696  - <code>dict\[str, Any\]</code> – A dictionary with the following keys:
697  - `replies`: A list of generated text completions as strings.
698  - `meta`: A list of metadata dictionaries containing information about each generation,
699    including model name, finish reason, and token usage statistics.